arXiv subjects

Tian Zhou

Publications and source records attributed to Tian Zhou.

At least 37 records · Page 2Linked to original sources

arXiv2026

Bridging Past and Future: Distribution-Aware Alignment for Time Series Forecasting

Although contrastive and other representation-learning methods have long been explored in vision and NLP, their adoption in modern time series forecasters remains limited. We believe they hold strong promise for this domain. To unlock this potential, we explicitly align past and future representations, thereby bridging the distributional gap between input histories and future targets. To this end, we introduce TimeAlign, a lightweight, plug-and-play framework that establishes a new representation paradigm, distinct from contrastive learning, by aligning auxiliary features via a simple reconstruction task and feeding them back into any base forecaster. Extensive experiments across eight benchmarks verify its superior performance. Further studies indicate that the gains arise primarily from correcting frequency mismatches between historical inputs and future outputs. Additionally, we provide two theoretical justifications for how reconstruction improves forecasting generalization and how alignment increases the mutual information between learned representations and predicted targets. The code is available at https://github.com/TROUBADOUR000/TimeAlign.

cs.LG↗

arXiv2026

Enhancing AI-Based Tropical Cyclone Track and Intensity Forecasting via Systematic Bias Correction

Tropical cyclones (TCs) pose severe threats to life, infrastructure, and economies in tropical and subtropical regions, underscoring the critical need for accurate and timely forecasts of both track and intensity. Recent advances in AI-based weather forecasting have shown promise in improving TC track forecasts. However, these systems are typically trained on coarse-resolution reanalysis data (e.g., ERA5 at 0.25 degree), which constrains predicted TC positions to a fixed grid and introduces significant discretization errors. Moreover, intensity forecasting remains limited especially for strong TCs by the smoothing effect of coarse meteorological fields and the use of regression losses that bias predictions toward conditional means. To address these limitations, we propose BaguanCyclone, a novel, unified framework that integrates two key innovations: (1) a probabilistic center refinement module that models the continuous spatial distribution of TC centers, enabling finer track precision; and (2) a region-aware intensity forecasting module that leverages high-resolution internal representations within dynamically defined sub-grid zones around the TC core to better capture localized extremes. Evaluated on the global IBTrACS dataset across six major TC basins, our system consistently outperforms both operational numerical weather prediction (NWP) models and most AI-based baselines, delivering a substantial enhancement in forecast accuracy. Remarkably, BaguanCyclone excels in navigating meteorological complexities, consistently delivering accurate forecasts for re-intensification, sweeping arcs, twin cyclones, and meandering events. Our code is available at https://github.com/DAMO-DI-ML/Baguan-cyclone.

cs.LG↗

arXiv2026

Baguan-TS: A Sequence-Native In-Context Learning Model for Time Series Forecasting with Covariates

Transformers enable in-context learning (ICL) for rapid, gradient-free adaptation in time series forecasting, yet most ICL-style approaches rely on tabularized, hand-crafted features, while end-to-end sequence models lack inference-time adaptation. We bridge this gap with a unified framework, Baguan-TS, which integrates the raw-sequence representation learning with ICL, instantiated by a 3D Transformer that attends jointly over temporal, variable, and context axes. To make this high-capacity model practical, we tackle two key hurdles: (i) calibration and training stability, improved with a feature-agnostic, target-space retrieval-based local calibration; and (ii) output oversmoothing, mitigated via context-overfitting strategy. On public benchmark with covariates, Baguan-TS consistently outperforms established baselines, achieving the highest win rate and significant reductions in both point and probabilistic forecasting metrics. Further evaluations across diverse real-world energy datasets demonstrate its robustness, yielding substantial improvements.

cs.LG↗

arXiv2025

SimDiff: Simpler Yet Better Diffusion Model for Time Series Point Forecasting

Diffusion models have recently shown promise in time series forecasting, particularly for probabilistic predictions. However, they often fail to achieve state-of-the-art point estimation performance compared to regression-based methods. This limitation stems from difficulties in providing sufficient contextual bias to track distribution shifts and in balancing output diversity with the stability and precision required for point forecasts. Existing diffusion-based approaches mainly focus on full-distribution modeling under probabilistic frameworks, often with likelihood maximization objectives, while paying little attention to dedicated strategies for high-accuracy point estimation. Moreover, other existing point prediction diffusion methods frequently rely on pre-trained or jointly trained mature models for contextual bias, sacrificing the generative flexibility of diffusion models. To address these challenges, we propose SimDiff, a single-stage, end-to-end framework. SimDiff employs a single unified Transformer network carefully tailored to serve as both denoiser and predictor, eliminating the need for external pre-trained or jointly trained regressors. It achieves state-of-the-art point estimation performance by leveraging intrinsic output diversity and improving mean squared error accuracy through multiple inference ensembling. Key innovations, including normalization independence and the median-of-means estimator, further enhance adaptability and stability. Extensive experiments demonstrate that SimDiff significantly outperforms existing methods in time series point forecasting.

cs.AI↗

arXiv2025

Guided MRI Reconstruction via Schrödinger Bridge

Magnetic Resonance Imaging (MRI) is an inherently multi-contrast modality, where cross-contrast priors can be exploited to improve image reconstruction from undersampled data. Recently, diffusion models have shown remarkable performance in MRI reconstruction. However, they still struggle to effectively utilize such priors, mainly because existing methods rely on feature-level fusion in image or latent spaces, which lacks explicit structural correspondence and thus leads to suboptimal performance. To address this issue, we propose $\mathbf{I}^2$SB-Inversion, a multi-contrast guided reconstruction framework based on the Schrödinger Bridge (SB). The proposed method performs pixel-wise translation between paired contrasts, providing explicit structural constraints between the guidance and target images. Furthermore, an Inversion strategy is introduced to correct inter-modality misalignment, which often occurs in guided reconstruction, thereby mitigating artifacts and improving reconstruction accuracy. Experiments on paired T1- and T2-weighted datasets demonstrate that $\mathbf{I}^2$SB-Inversion achieves a high acceleration factor of up to 14.4 and consistently outperforms existing methods in both quantitative and qualitative evaluations.

eess.IV↗

arXiv2025

A Spin-Based Pathway to Testing the Quantum Nature of Gravity

A key open problem in physics is the correct way to combine gravity (described by general relativity) with everything else (described by quantum mechanics). This problem suggests that general relativity and possibly also quantum mechanics need fundamental corrections. Most physicists expect that gravity should be quantum in character, but gravity is fundamentally different to the other forces because it alone is described by spacetime geometry. Experiments are needed to test whether gravity, and hence space-time, is quantum or classical. We propose an experiment to test the quantum nature of gravity by checking whether gravity can entangle two micron-sized crystals. A pathway to this is to create macroscopic quantum superpositions of each crystal first using embedded spins and Stern-Gerlach forces. These crystals could be nanodiamonds containing nitrogen-vacancy (NV) centres. The spins can subsequently be measured to witness the gravitationally generated entanglement. This is based on extensive theoretical feasibility studies and experimental progress in quantum technology. The eventual experiment will require a medium-sized consortium with excellent suppression of decoherence including vibrations and gravitational noise. In this white paper, we review the progress and plans towards realizing this. While implementing these plans, we will further explore the most macroscopic superpositions that are possible, which will test theories that predict a limit to this.

quant-ph↗

arXiv2025

Gyroscopic stability for nanoparticles in Stern-Gerlach Interferometry and spin contrast

Creating macroscopic spatial quantum superposition with a nanoparticle has a multitude of applications, ranging from testing the foundations of quantum mechanics, matter-wave interferometer for detecting gravitational waves and probing the electromagnetic vacuum, dark matter detection and quantum sensors to testing the quantum nature of gravity in a lab. In this paper, we investigate the role of rotation in a matter-wave interferometer, where we show that imparting angular momentum along the direction of a defect, such as one present in the nitrogen-vacancy centre of a nanodiamond can cause an enhancement in spin contrast for a wide-ranging value of the angular momentum, e.g. $10^{3}-10^{6}$~Hz for a mass of order $10^{-14}-10^{-17}$ Kg nanodiamond. Furthermore, the imparted angular momentum can enhance the spatial superposition by almost a factor of two and possibly average out any potential permanent dipoles in the nanodiamond.

quant-ph↗

arXiv2025

Spin contrast, finite temperature, and noise in matter-wave interferometer

In this paper, we will show how finite-temperature corrections and spin-dependent/independent noise will affect the contrast in a matter-wave interferometer, especially with massive objects and large spatial superposition sizes. Typically, spin is embedded in a nanoparticle as a defect, which can be manipulated by the external magnetic field to create a macroscopic quantum superposition. These massive matter-wave interferometers are the cornerstone for many new fundamental advancements in physics; particularly, macroscopic quantum superposition can use entanglement features to, e.g., test physics beyond the Standard Model, test the equivalence principle, improve quantum sensors, and test the quantum nature of spacetime in a lab. We will consider a Stern-Gerlach type apparatus to create macroscopic quantum superposition in a harmonic oscillator trap, and figure out the spin contrast loss due to linear spin-independent and spin-dependent noise in a single interferometer. We will show that spin contrast loss due to spin-independent noise does not depend on the initial thermal state of the matter wave function. However, spin contrast loss due to spin-dependent fluctuations do depend on the initial thermal occupation of the quantum state. We will keep our discussion general as far as the noise parameters are concerned.

quant-ph↗

arXiv2025

Rotational stability in nanorotor and spin contrast in one-loop interferometry in the Stern-Gerlach setup

The rotation of a nanoparticle in a quantum system has many applications, from theory to experiments. This paper will treat nanoparticle rotational dynamics for spin-embedded nanorotors. We will model it as a rigid body that properly treats the rotation in the co-frame of the nanorotor in the presence of external fields. Besides rotation, we will further investigate how to create large spatial superpositions in the inhomogeneous external magnetic field, such as in the case of the Stern-Gerlach apparatus. The spin-embedded nanorotors play a crucial role in creating matter-wave interferometers through their spin and external magnetic field interaction Hamiltonian. We aim to provide a holistic interpretation of the dynamics of three Euler angles, their quantum evolution, and the nanorotor's spatial motion in a Stern-Gerlach-type setup where we will consider one-full-loop interferometry. We will then study how the quantum evolution of all the Euler angles leads to a spin coherence loss upon interference and what manifests the Einstein-de Haas effect in an external magnetic field. In particular, we show that by imparting rotation along the direction of the magnetic field, we can stabilise the nanorotor's libration mode. We will also extend our analysis to a case where the initial state of the libration mode is thermal and discuss the contrast loss due to interference of the nanorotor upon one-loop completion.

quant-ph↗

arXiv2025

Utilizing Strategic Pre-training to Reduce Overfitting: Baguan -- A Pre-trained Weather Forecasting Model

Weather forecasting has long posed a significant challenge for humanity. While recent AI-based models have surpassed traditional numerical weather prediction (NWP) methods in global forecasting tasks, overfitting remains a critical issue due to the limited availability of real-world weather data spanning only a few decades. Unlike fields like computer vision or natural language processing, where data abundance can mitigate overfitting, weather forecasting demands innovative strategies to address this challenge with existing data. In this paper, we explore pre-training methods for weather forecasting, finding that selecting an appropriately challenging pre-training task introduces locality bias, effectively mitigating overfitting and enhancing performance. We introduce Baguan, a novel data-driven model for medium-range weather forecasting, built on a Siamese Autoencoder pre-trained in a self-supervised manner and fine-tuned for different lead times. Experimental results show that Baguan outperforms traditional methods, delivering more accurate forecasts. Additionally, the pre-trained Baguan demonstrates robust overfitting control and excels in downstream tasks, such as subseasonal-to-seasonal (S2S) modeling and regional forecasting, after fine-tuning.

cs.LG↗

arXiv2025

Output Scaling: YingLong-Delayed Chain of Thought in a Large Pretrained Time Series Forecasting Model

We present a joint forecasting framework for time series prediction that contrasts with traditional direct or recursive methods. This framework achieves state-of-the-art performance for our designed foundation model, YingLong, and reveals a novel scaling effect: longer outputs significantly enhance model accuracy due to delayed chain-of-thought reasoning in our non-causal approach. YingLong is a non-causal, bidirectional attention encoder-only transformer trained through masked token recovery, aligning more effectively with language understanding tasks than with generation tasks. Additionally, we boost performance by tackling output variance with a multi-input ensemble. We release four foundation models ranging from 6M to 300M parameters, demonstrating superior results in zero-shot tasks on the ETT and Weather datasets. YingLong achieves more than 60% best performance. To ensure generalizability, we assessed the models using the GIFT-Eval benchmark, which comprises 23 time series datasets across 7 domains. Yinglong significantly outperformed the best time-series foundation models, end-to-end trained models by 14% and 44% in rank respectively.The pretrained 300M model is available at https://huggingface.co/qcw1314/YingLong_300m

cs.LG↗

arXiv2025

Does Vector Quantization Fail in Spatio-Temporal Forecasting? Exploring a Differentiable Sparse Soft-Vector Quantization Approach

Spatio-temporal forecasting is crucial in various fields and requires a careful balance between identifying subtle patterns and filtering out noise. Vector quantization (VQ) appears well-suited for this purpose, as it quantizes input vectors into a set of codebook vectors or patterns. Although VQ has shown promise in various computer vision tasks, it surprisingly falls short in enhancing the accuracy of spatio-temporal forecasting. We attribute this to two main issues: inaccurate optimization due to non-differentiability and limited representation power in hard-VQ. To tackle these challenges, we introduce Differentiable Sparse Soft-Vector Quantization (SVQ), the first VQ method to enhance spatio-temporal forecasting. SVQ balances detail preservation with noise reduction, offering full differentiability and a solid foundation in sparse regression. Our approach employs a two-layer MLP and an extensive codebook to streamline the sparse regression process, significantly cutting computational costs while simplifying training and improving performance. Empirical studies on five spatio-temporal benchmark datasets show SVQ achieves state-of-the-art results, including a 7.9% improvement on the WeatherBench-S temperature dataset and an average mean absolute error reduction of 9.4% in video prediction benchmarks (Human3.6M, KTH, and KittiCaltech), along with a 17.3% enhancement in image quality (LPIPS). Code is publicly available at https://github.com/Pachark/SVQ-Forecasting.

cs.CV↗

arXiv2025

Unlocking the Power of LSTM for Long Term Time Series Forecasting

Traditional recurrent neural network architectures, such as long short-term memory neural networks (LSTM), have historically held a prominent role in time series forecasting (TSF) tasks. While the recently introduced sLSTM for Natural Language Processing (NLP) introduces exponential gating and memory mixing that are beneficial for long term sequential learning, its potential short memory issue is a barrier to applying sLSTM directly in TSF. To address this, we propose a simple yet efficient algorithm named P-sLSTM, which is built upon sLSTM by incorporating patching and channel independence. These modifications substantially enhance sLSTM's performance in TSF, achieving state-of-the-art results. Furthermore, we provide theoretical justifications for our design, and conduct extensive comparative and analytical experiments to fully validate the efficiency and superior performance of our model.

cs.LG↗

arXiv2025

Towards Expressive Spectral-Temporal Graph Neural Networks for Time Series Forecasting

Time series forecasting has remained a focal point due to its vital applications in sectors such as energy management and transportation planning. Spectral-temporal graph neural network is a promising abstraction underlying most time series forecasting models that are based on graph neural networks (GNNs). However, more is needed to know about the underpinnings of this branch of methods. In this paper, we establish a theoretical framework that unravels the expressive power of spectral-temporal GNNs. Our results show that linear spectral-temporal GNNs are universal under mild assumptions, and their expressive power is bounded by our extended first-order Weisfeiler-Leman algorithm on discrete-time dynamic graphs. To make our findings useful in practice on valid instantiations, we discuss related constraints in detail and outline a theoretical blueprint for designing spatial and temporal modules in spectral domains. Building on these insights and to demonstrate how powerful spectral-temporal GNNs are based on our framework, we propose a simple instantiation named Temporal Graph Gegenbauer Convolution (TGGC), which significantly outperforms most existing models with only linear components and shows better model efficiency. Our findings pave the way for devising a broader array of provably expressive GNN-based models for time series.

cs.LG↗

arXiv2024

Maximizing the Impact of Deep Learning on Subseasonal-to-Seasonal Climate Forecasting: The Essential Role of Optimization

Weather and climate forecasting is vital for sectors such as agriculture and disaster management. Although numerical weather prediction (NWP) systems have advanced, forecasting at the subseasonal-to-seasonal (S2S) scale, spanning 2 to 6 weeks, remains challenging due to the chaotic and sparse atmospheric signals at this interval. Even state-of-the-art deep learning models struggle to outperform simple climatology models in this domain. This paper identifies that optimization, instead of network structure, could be the root cause of this performance gap, and then we develop a novel multi-stage optimization strategy to close the gap. Extensive empirical studies demonstrate that our multi-stage optimization approach significantly improves key skill metrics, PCC and TCC, while utilizing the same backbone structure, surpassing the state-of-the-art NWP systems (ECMWF-S2S) by over \textbf{19-91\%}. Our research contests the recent study that direct forecasting outperforms rolling forecasting for S2S tasks. Through theoretical analysis, we propose that the underperformance of rolling forecasting may arise from the accumulation of Jacobian matrix products during training. Our multi-stage framework can be viewed as a form of teacher forcing to address this issue. Code is available at \url{https://anonymous.4open.science/r/Baguan-S2S-23E7/}

cs.LG↗

arXiv2024

Understanding the Role of Textual Prompts in LLM for Time Series Forecasting: an Adapter View

In the burgeoning domain of Large Language Models (LLMs), there is a growing interest in applying LLM to time series forecasting, with multiple studies focused on leveraging textual prompts to further enhance the predictive prowess. This study aims to understand how and why the integration of textual prompts into LLM can effectively improve the prediction accuracy of time series, which is not obvious at the glance, given the significant domain gap between texts and time series. Our extensive examination leads us to believe that (a) adding text prompts is roughly equivalent to introducing additional adapters, and (b) It is the introduction of learnable parameters rather than textual information that aligns the LLM with the time series forecasting task, ultimately enhancing prediction accuracy. Inspired by this discovery, we developed four adapters that explicitly address the gap between LLM and time series, and further improve the prediction accuracy. Overall,our work highlights how textual prompts enhance LLM accuracy in time series forecasting and suggests new avenues for continually improving LLM-based time series analysis.

cs.LG↗

arXiv2024

Effect of near-earth thunderstorm electric field on the flux of cosmic ray air showers in LHAASO-KM2A

The Large High Altitude Air Shower Observatory (LHAASO) is located at Haizi Mountain, Daocheng, Sichuan province, China. Due to its high-altitude location with frequent thunderstorm activities, the LHAASO is suited for studying the effects of near-earth thunderstorm electric fields on cosmic ray air showers. In this paper, Monte Carlo simulations are performed with CORSIKA and G4KM2A to analyze the flux variations of cosmic ray air showers detected by the kilometer-square array of LHAASO (LHAASO-KM2A) during thunderstorms. The strength, polarity, and layer thickness of atmospheric electric field (AEF) during thunderstorm are found to be associated with the shower rate variations. The flux of shower events satisfying trigger conditions of the KM2A increases with field intensity, particularly within negative fields, and the enhanced amplitude is more than 5% in -600 V/cm and 12% in -1000 V/cm, whereas it increases by only 1% and 7% in equivalent positive fields, respectively. While in positive fields ranging from 0 to 400 V/cm, the shower rate decreases with smaller amplitudes. Furthermore, the shower rate increases dramatically with the AEF layer thickness until a certain value, above which the variation trend slows down. The dependence of the trigger rate variation on the primary zenith angle has also been revealed, increasing in lower zenith angle ranges and showing opposite behaviors in higher ones. Additionally, we study that the relationship between the trigger rate variations and the primary energies, and find the enhanced amplitude of the shower rate decreases with increasing primary energy. Simultaneously, the shower events with lower primary energy show a significant increase, whereas events with higher primary energy are hardly affected during thunderstorms. Our simulations offer insights into the variation of the trigger rate detected by LHAASO-KM2A during thunderstorms.

astro-ph.HE↗

arXiv2024

Mitigating Time Discretization Challenges with WeatherODE: A Sandwich Physics-Driven Neural ODE for Weather Forecasting

In the field of weather forecasting, traditional models often grapple with discretization errors and time-dependent source discrepancies, which limit their predictive performance. In this paper, we present WeatherODE, a novel one-stage, physics-driven ordinary differential equation (ODE) model designed to enhance weather forecasting accuracy. By leveraging wave equation theory and integrating a time-dependent source model, WeatherODE effectively addresses the challenges associated with time-discretization error and dynamic atmospheric processes. Moreover, we design a CNN-ViT-CNN sandwich structure, facilitating efficient learning dynamics tailored for distinct yet interrelated tasks with varying optimization biases in advection equation estimation. Through rigorous experiments, WeatherODE demonstrates superior performance in both global and regional weather forecasting tasks, outperforming recent state-of-the-art approaches by significant margins of over 40.0\% and 31.8\% in root mean square error (RMSE), respectively. The source code is available at \url{https://github.com/DAMO-DI-ML/WeatherODE}.

cs.LG↗